clud-bug-collaboration
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and act upon data from external PR review comments generated by the 'clud-bug' bot. This creates a potential surface for indirect prompt injection if the bot's output is compromised.
- Ingestion points: The agent reads PR review comments (v0.6.5+) and the
.clud-bug.jsonconfiguration file. - Boundary markers: The skill specifies a structured finding shape (e.g., triage lines and emoji prefixes like 🔴, 🟡, 🟣), which provides structural delimitation, though it lacks explicit 'ignore instructions' warnings for the content.
- Capability inventory: The agent is authorized to edit repository files, address review comments, and push commits based on the bot's findings.
- Sanitization: The instructions do not define specific sanitization or escaping protocols for the content retrieved from the review comments.
- [SAFE]: The skill documentation includes references to the
clud-bugCLI and theanthropics/claude-code-action. These resources originate from the skill author's own organization and a well-known service, respectively, and are used for their intended primary purpose of repository management and PR review.
Audit Metadata